Teen Developer Releases AI-Powered Tech Debt Audit Tool for Local Repos

As a teenager learning to code, I once hesitated before sharing my project with experienced developers—not because I doubted the tool, but because I feared their reaction to my code. That fear didn’t stop me from shipping Tech Debt Agent, a provider-agnostic CLI tool that audits repositories for technical debt, refactors files using AI, and generates a local dark-mode dashboard without ever exposing your code to the internet.

Behind the Tool: Solving a Universal Developer Pain Point

Technical debt isn’t just a problem for legacy systems—it exists in every codebase, from solo projects to enterprise repositories. Files grow unchecked, variables become cryptic, and functions balloon beyond reason. Most tools either produce superficial linting reports, require invasive cloud uploads, or demand complex setup before delivering value.

What if a tool could act like a meticulous senior engineer? One that doesn’t just point out problems but quantifies them, fixes them in a safe environment, and presents findings in a clear, actionable way? That’s what motivated me to build Tech Debt Agent.

How Tech Debt Agent Works: Three Core Components

1. The Auditor: Quantifying Mess Before It Becomes Crisis

When you point the tool at a directory, it performs a deep scan for:

• Security risks like hardcoded secrets or unsafe eval usage
• Structural issues such as deeply nested logic or duplicated code blocks
• Maintenance burdens like dead code or overly complex functions

Instead of vague warnings, the tool calculates Debt Hours—an estimate of the time a developer would need to clean up each file. This transforms subjective perceptions of mess into prioritized, data-driven insights.

2. The Sandbox Refactor: AI Improvements Without Risk

The most defining feature of Tech Debt Agent is its commitment to safety. The tool never modifies your original code. Instead, it creates a quarantine folder called fixed_code_output/. Every refactored file lands here, complete with explanations of what changed and why. You can:

• Compare original and refactored versions
• Adopt only the improvements you trust
• Discard the output entirely if needed

This approach eliminates the fear of automated changes derailing a project.

3. The Dashboard: Visualizing Debt for Actionable Insights

After scanning, the tool generates a single-file, dark-mode HTML dashboard that opens automatically in your browser. No frameworks, no servers, no internet connection required. The dashboard displays:

• Debt scores per file
• Breakdowns by debt type (security, logic, maintenance)
• Rankings of the worst offenders
• Overall repository health metrics

It turns raw data into a shareable, archivable report your entire team can use.

Privacy-First Design: Your Code Stays Yours

For professional developers and companies, privacy isn’t optional—it’s essential. Tech Debt Agent supports multiple inference providers:

• OpenAI or Anthropic APIs for convenience
• Ollama for fully offline, local inference
• LM Studio for a GUI-based local model setup

Running with Ollama ensures zero bytes of your code leave your machine. For teams handling proprietary code, this isn’t a feature—it’s a necessity.

Getting Started in Three Simple Steps

git clone 
cd tech-debt-agent
pip install -e .

Run an audit on any directory:

techdebt audit ./your-project

Or execute it entirely offline with Ollama:

techdebt audit ./your-project --provider ollama --model codellama

No configuration files. No account creation. Just immediate value.

Lessons Learned From Shipping a Production Tool at 16

Sandboxing was the hardest but most crucial decision.

My initial prototype modified files directly. After sharing it with a developer who flatly refused to use it, I realized the tool needed a quarantine approach. The fixed_code_output/ folder emerged from that harsh feedback.

Provider-agnostic design forced better engineering.

Supporting cloud APIs, Ollama, and LM Studio meant building a clean abstraction layer early. The result? More modular, maintainable Python than I’d ever written before.

The dashboard made the tool feel real.

CLI tools with colored text are useful—but they don’t feel like products. The moment I added the visual dashboard, Tech Debt Agent transitioned from a script to something teams could genuinely rely on.

Your Turn: Break It, Use It, Improve It

I’m not asking for stars or shares. I’m asking you to run Tech Debt Agent on your messiest directory and tell me:

• What broke?
• What’s missing?
• What could be smarter?
• What did you actually like?

Open a GitHub issue, leave a comment, or reach out directly. Every piece of feedback will shape the next version.

I’m 16. I’m learning in public. And I built something I genuinely believe solves a real problem. Prove me right. Or prove me wrong. Either way, I want to know.